Just-in-time learning Enhanced Huber Regression Model for Flow Rate Sensor Anomaly Detection in Propellant Filling Process

Flow rate is a key measurement parameter in the process of industrial production and operation, especially in high safety-critical scenarios such as the propellant filling process of space launch vehicles. In the face of special environments such as strong corrosion, inflammable and explosive, and complex conditions such as multi-stage simultaneous filling, redundant parallel operation of equipment, and reflux, timely and accurate abnormal flow detection is crucial to ensure the safe and stable operation of the system. Aiming at the nonlinear characteristics and dynamic conditions of propellant filling process of a space launch vehicle, a just-in-time learning enhanced Huber regression soft-sensing model is proposed in this paper. By integrating sensor data such as valve opening, current and pressure, flow parameters can be accurately estimated for sensor anomaly detection. The effectiveness and robustness of the proposed method are verified on the actual data of 20 filling processes, which provides a feasible technical scheme for the safety control of propellant filling system of a space launch vehicle.